The article discusses a study that aims to improve virtual sound synthesis, specifically the generation of head-related impulse responses (HRIRs). These responses are crucial for creating accurate spatial sound experiences through headphones. However, they are difficult to measure due to individual differences and the need for specialized environments. The researchers propose using temporal convolutional neural networks (TCNs) to generate HRIRs from one direction to another. The TCNs were trained using publicly available datasets and were successful in generating HRIRs for different directions. Although there was some degradation in spectral distortion, the generated HRIRs were found to be equivalent to measured ones. This suggests that the proposed method could be used for personalizing virtual sound.
Publication date: 25 Oct 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2310.14018
